We ought to test Elixir, but most of the human time spent on it is related to performance issues more than bugs per se. Our testing infrastructure should include some benchmarking facilities to easily compare two Elixir commits. Eg:
- Various pages response time. At least one per request handler.
- Indexing (without taking into account cloning of Git repos).
- Indexing of a 0-new-blob tag (seems slow currently, can be done by creating a/more new tag using
for i in $(seq $N); do git tag v100.$i $some_commit; done. Adding 6950 tags to musl (already indexed) and reindexing took: wall 52.63m, usr 595.03s, sys 511.72s. Done on 2609780.
- etc.
Not high priority.
We ought to test Elixir, but most of the human time spent on it is related to performance issues more than bugs per se. Our testing infrastructure should include some benchmarking facilities to easily compare two Elixir commits. Eg:
for i in $(seq $N); do git tag v100.$i $some_commit; done. Adding 6950 tags to musl (already indexed) and reindexing took: wall 52.63m, usr 595.03s, sys 511.72s. Done on 2609780.Not high priority.